Properly designing scattered pilot pattern over orthogonal frequency division multiplexing (OFDM) subcarriers is important to improve the accuracy of the sparse channel estimation, while decreasing the number of the required pilots. In this paper, we propose a pilot-subcarrier allocation scheme that optimizes the pilot subcarrier patterns without any knowledge of channels, where channel estimation is done by interpolating the pilot- subcarriers scattered over frequency domain. The proposed pilot allocation scheme utilizes the mutual incoherence property (MIP) of the compressive sensing (CS) theory to formulate the pilot allocation problem into infinity norm problem. Then, MIP based weighted fast iterative shrinkage-thresholding algorithm (MIP-WFISTA) is proposed to solve the formulated problem. Simulation results validate that, compared with the conventional techniques, the proposed pilot design scheme achieves more accurate channel estimation and as a result better bit error rate (BER) performance while decreasing the number of the required pilots in frequency-selective fading environments.